How to Vet an Automation Consultant

Why the Usual Signals Don't Work Here

Automation is an easy field to look competent in and a hard one to be good at. The tools are visual, the demos are quick, and a workflow that works on clean test data can be built by almost anyone in an afternoon.

That means the normal hiring signals underperform. Certifications mostly prove someone completed a vendor's course. Portfolios show finished canvases, which reveal nothing about whether they still run. Testimonials are collected at handover, which is exactly when every automation works.

What actually predicts a good engagement is whether the person has operated automations over time, not just built them. These questions get at that.

The Six Questions

"What happens when this breaks?"

The single most useful question, and we'd ask it first. A builder will describe the workflow again. An operator will talk about retries, timeouts, what escalates to a human, and how you find out. If failure hasn't crossed their mind, everything they build for you will fail silently, because that's the default behaviour of every automation platform.

"How will I know it stopped working?"

Related and distinct. Errors are easy to catch. A scheduled job that never fires produces no error at all, and that's the failure that costs money because nothing is there to notice. Listen for whether they mention heartbeats or absence-detection unprompted.

"Whose accounts will this run on?"

The answer should be yours. Your Twilio, your Google Workspace, your automation platform, billed to you, with credentials in a manager you control. If the build lives in their workspace or authenticates through their personal logins, you are renting something you think you're buying, and the invoice for that arrives the day you part company.

"What would you tell me not to automate?"

Anyone experienced has a list. Processes that change monthly. Steps where two people disagree about ownership. Anything with volume too low to repay the build. A consultant who thinks your whole business is a candidate is selling hours.

"Can I see a workflow you built, not a case study?"

Thirty seconds looking at an actual canvas tells you more than a page of results. You're looking for named steps, visible error branches, and logic you can roughly follow without being an expert. A wall of unnamed HTTP nodes and inline scripts is a maintenance bill.

"What happens if we stop working together?"

Documentation, access transfer, and whether the thing keeps running. Ask it early. The reaction is informative on its own.

Warning Signs

They quote before understanding the process. A number produced from a two-line brief is a guess, and it'll be revised once the work starts.

No questions about your data. Data quality drives more of the cost and risk than the workflow logic does. Someone who never asks how clean your records are hasn't hit that wall yet, or is planning to bill for it later as a surprise.

Everything needs their platform. Some shops have built a proprietary layer and every problem routes through it. Fine when it genuinely fits. Worth knowing that you're buying a dependency with a subscription.

AI on every step. Covered properly in what to look for in an AI automation consultant, and the short version is that a model belongs in the few places where inputs are genuinely unstructured and nowhere near anything with a right answer.

Retainer required from day one. Before anything exists to maintain, that's a subscription rather than a service.

No mention of handover. If nobody plans for you to understand what was built, you'll be calling them for every change.

A Cheap Way to Test Someone

Before committing to a large engagement, buy one small, well-defined workflow. A single automation with a clear before and after, at a price you'd be willing to lose.

You learn things a proposal cannot tell you. Whether they ask good questions during scoping. Whether they build error handling without being asked, or only after you raise it. Whether the documentation arrives. Whether they hit the date. Whether the thing is still running in six weeks without anyone touching it.

That last one is the real test, and it's why we'd suggest waiting a month before deciding on the bigger piece of work. Plenty of automations look finished on Friday and quietly stop by the middle of the following month.

This also matches the sequencing that works best regardless of who builds it, which we covered in automate one workflow first, not a platform.

Questions They Should Be Asking You

The conversation runs both ways, and a good consultant will want to know things you might not have thought about.

How many records or transactions go through this per week, because volume changes the design. Who owns each step today. What happens to the process when the person who normally does it is on holiday. Whether anything here touches money, health data or anything regulated. What you've already tried and why it didn't stick.

If nobody asks about volume or about what's been tried before, you're being scoped generically.

Making the Decision

Weight actual operating experience above tool credentials, and weight how someone talks about failure above how they talk about capability. Every automation platform can do the happy path. What you're paying for is the judgement about everything either side of it.

If you're still deciding what type of provider fits, what an automation consultant actually does covers the four kinds, what automation consulting actually costs covers pricing, and workflow automation services: what you get is the delivery checklist to hold any provider against.

Book a free 60-minute call. Ask us all six questions. If our answers don't hold up, you've lost an hour and gained a scope you can take elsewhere.